| |
Interview with Colin Green
Pioneer of Previsualization and founder of Pixel Liberation Front Colin Green talks about the early days of cg, modern filmmaking, the importance of previs and how it all began. Credits include Fight Club, Panic Room, Eraser, Matrix Reloaded, Matrix Revolutions, Minority Report and Superman Returns among others.
April, 1st, 2007, by Raffael Dickreuter
 |  | Colin Green, Pioneer of Previs and
Founder of Pixel Liberation Front.
| | | |
| |
| | | |
|
Tell us about how it was when you were starting out and ended up breaking into the film industry.
My background prior to working visual effect was in Architecture. I did an undergraduate major in Architecture at Yale, and was one year into a Master's program at MIT when I got a job with Doug Trumbull's company in the Berkshires. I had a focus on CAD and 3D modeling as a design tool. I cross-registered to the Harvard Graduate School of Design to take a 3D modeling course, taught by William Mitchell, which focused on instruction in the original Alias software, I think version 3. I remember almost falling over the first time I saw the 3D viewport "orbit" an object. CAD software available then on PCs took forever to draw a single view of a 3D object, so being able to spin freely around it was like magic. I managed to get some time one of Harvard's two SGI 4D workstations, mostly after midnight, guiding myself through the software. It was quite different from the solid modeling CAD systems I was used to and I am sure most of what I did was completely wrong. No one really knew how it was intended to be used back then.
Barely knowing Alias (or Unix), a bunch of us went off to work for Doug Trumbull, where, faced with real-world tasks, we learned what we needed on the job by spending hour after hour on the machines. One guy in our group, Terrence Masson, knew how to use the software pretty well, and he helped a lot in getting us to turn out usable work. (Terrence chaired last year's SIGGRAPH Electronic Theater.)
What do you like to do in your spare time?
I enjoy many outdoor activities -- hiking, mountain biking, snowboarding, rock climbing, mountaineering, kayaking, none of which I have as much time for as I'd like while working in VFX. Traveling is also a favorite, and being employed in visual effects and various overseas projects has been helpful in maintaining this particular interest.
How was it working with Douglas Trumbull and in times when computers and cg software were not that advanced as today?
Doug had been hired by Circus Circus Enterprises to create and produce three "special venue" attractions for the company's new Luxor Hotel and Casino project. These projects were intended to be the most spectacular, technically advanced VFX experiences ever created, with a total running time of about 15 minutes and a budget of something like $50M.
It turned out to be the best possible first job scenario for me due to the fact that it encompassed every aspect of vfx production available at the time and was aiming to set a new benchmark. Rather than using CG to create spinning logos for a client, I was able to participate in a one-of-a-kind integration of all available VFX techniques. Doug had promised the most advanced and cutting-edge show ever created, and brought in many top people in their fields. Cameras, projectors, lenses, motion control rigs, lighting gear were all custom manufactured, and were the best that could be made. You can imagine the excitement of being part of the lunchroom conversations as the whole process was being mapped out.
Doug had made his name on 2001: Space Odyssey by pioneering the Slit Scan process at age 21, and winning an Oscar. Since it had worked for him, he was very comfortable empowering very smart young people with amazing authority over creative problem-solving and technological pioneering. Being a part of a process with this rare mindset exciting as well as fun.
Almost everything was new to almost everyone, at least at this level of realization and integration with other disciplines. Computer tools and methodologies were being developed and incorporated for the first time on this scale for most people involved. This was a few years before Jurassic Park, and the CG industry was just beginning to pick up steam. Jeff Kleiser and Diana Walczak had their company, KWCC, which was setting up to do all the CGI elements for the project using Wavefront. The most experienced CG people had mainly done commercial work at video res. Film res digital effects hardly existed.
This was a totally different era of VFX, where even digital compositing for film was not yet really "done" except through extremely proprietary pipelines. The computer that was used to do the film res compositing (an IBM PVS) was almost one of a kind, and driven by Serge Sretchinski, who wrote the software, the operating system, composited all the shots, almost completely on his own, in a tiny dark and overheated room upstairs... Talk about Black Arts.
Those of us from Harvard and MIT Architecture worked in a department called "Image Engineering," which was intended as a front end for precision design and fabrication of miniature sets. The idea was to provide CAD/CAM data to the Modelshop so that the miniatures, once built, would be would be seen by the motion control camera on the set exactly the same as they were in 3D in the computer. Using our SGI machines for this task proved to be time consuming and not really very useful, as the model builders were perfectly capable of building accurate models from paper plans with normal table saws and exacto knives. There were some embarrassing failures, and the costs were not cheap.
Not so well regarded by the traditional model building and motion control folks and too inexperienced to compete with what KWCC were doing in CG, we were misfits. But misfits who had been placed at the center of a highly complex and untested production process and who really wanted to keep going to find a way to make a meaningful contribution and to define our role.
Once the film portion was done, I moved into ride programming. Continuing the idea of using 3D animation tools to better integrate with real, physical aspects of the show, I developed an interface allowing me to use Softimage to design and execute the movement of a hydraulic motion base for the ride. I spent five months in Las Vegas doing the installation of the show, and final tweaks of the ride until it opened. So a job which was intended to be over in six months months lasted over two years.
the early days.
You are considered as one of the inventors of the modern previs process. Tell us how it all started and why this actually happened.
I first used the term "Previsualization" following the Luxor project. The studio was closed up for months, and eventually sold to Andy Vajna's Cinergi productions, which was producing "Judge Dredd." As this project required large miniature sets integrated with CGI, our studio was a natural fit and, led by Joel Hynek, the whole company was reformulated. I was asked by Joel to take over the lead of Image Engineering. I changed the name of the department to Previsualization and gave up most of the miniature fabrication duties to focus on shot design and technical planning.
One of the first decisions I made as I took over the department was switching from Alias to Softimage 3D, which was simpler, faster and nicer to work with for animation. I have been solidly committed to using Softimage ever since, although the emergence of Maya as the dominant package has made this more difficult over time.
John Gaeta was very influential in making this process work well as a viable and respectable planning tool for all departments and across the whole of production. Having learned from our failures on Luxor, I was passionate about proving that the process could work, and be an important crossover point between motion control miniatures and CG, and tie the whole process together. We largely achieved this goal on Dredd, although it was not perfect yet.
As an example of this, we used something we called "witness point tracking" and surveys of markers to track a helicopter element of a swimming pool, and use this data to drive the motion control camera to shoot additional passes of miniature environments to extend this element, placing the pool on top of a futuristic sky scraper at the top of a vast MegaCity. This somehow worked and was considered to be extremely advanced and cutting edge, involving all original code (written by Serge).The effectiveness of what we did for Dredd confirmed the new promise of using 3D graphics in the production of integrated miniature photography and CGI and dispelled criticism that this work might just be cool to look at and not really at the heart of the design and technical execution process.
One final achievement, which was fun and a big step forward for us, was that the Previs department improved the look of our pre-vis city, and used these buildings to create a final film-res vista, filling the horizon of the most complex shot. We used a beta version of mental ray to render these elements, and developed a custom atmosphere shader.
Miniature set for Judge Dredd.
Previs for 'Judge Dredd'.
How did your career lead towards PLF getting started
When The Trumbull Company completed Luxor, the studio ceased operation and locked its doors, leaving us literally out in the cold of the Berkshires winter and scraping to cover the next month's rent. We were powerless to do anything but wait for another job; Kleiser Walczak, the CGI wing of the company, on the otherhand owned all of its machines and continued to operate (and make money) from its section of the building. I was inspired this example and by the advantages of being a business owner instead of an employee, and vowed never be in that situation again. Supported by my college friend Andrew Kress, I founded PLF. I also invited the 3 most talented and promising individuals I had worked with in my department on Dredd to be partners in the company. They later went on to set up Giant Killer Robots in San Franscisco, and have done very well. We bought one SGI indy machine and a Softimage license. We scraped together our savings and bought one SGI indy machine and one Softimage license, opened a small office in NYC, and we were in business.
What is the importance of Previs in todays process of filmmaking and how is it used on big productions?
Previs has come a long way since Judge Dredd. In selling our services to films, I used to have to make elaborate arguments about how we could help make vfx production more efficient, and how we could save money, etc. Now, film producers call and simply say "we need previs."
Previs was intended to be -- and works best as -- a cinematic design tool for the director and his/her key creative crew. As familiarity with Previs has expanded, so has the audience, and the participants now include many more people. On some projects, the pre-vis process seems to be most valued by studio personnel who want to be able to "see" the movie they are paying for earlier in the process, and approve budgets based on these views. This is certainly a valid evolution of pre-vis, but for me, it becomes less valid the farther away it is from being the tool the director uses to work out and define his/her vision.
Miniature set for Judge Dredd.
How were the early years when previs was still a new approach. Did lots of jobs start to pour in or did it take time to get directors used to it?
We were fortunate to find a few important early supporters. Scott E. Anderson, who was VFX Supervisor at Imageworks for Starship Troopers, was helpful in making PLF an important part of their miniature spaceship photography production pipeline, and we were able to greatly streamline the design and execution process for that complex project. We even stayed on after the shoot was over, and did preliminary shot integration, precomps, and layout for CG elements. Starship Troopers was key in that it showed me that a small company with few resources but with the right approach and right people could still be valuable to a major project. We put in a lot of long hours, but it paid off hugely in confidence and contacts. It also paved the way for us to move what was then still a NY-based company to L.A.
Previs for 'Starship Troopers'.
Luxor team photo. Colin is in the front row on the left (blue shirt). Right next to him in sunglasses John Gaeta.
How has it changed over the years if you look back and compare it todays standards.
From a business perspective, the cost structure for doing the work has changed radically. SGI machines cost $20,000 to $75,000 in the early days, and a license of Softimage was $50,000. We just had to charge enough to pay the lease costs. Paying for the gear was the biggest concern in the beginning, and getting more jobs meant we could get more gear. Today, costs for people are the largest expense. And you have to find lots of work to keep really good people on between big jobs.
From a creative/ artistic standpoint, the tool set now is so far beyond what we had back then. We did not do many aspects of the process simply because the tools did not exist. Editing, for example, could only really happen in an Avid; there was no low-end alternative. There was no After Effects. Even inside Softimage, which was by far the most advanced tool of the time, many aspects of CG technique were out of reach for Pre-vis, being to complex to implement. Our first fully enveloped character rig was not implemented until XSI v 3.5, and until then, we had separate geometry for all the limbs of a character, mostly because the rigging process was too complex and slow for Previs.
How did features handle complex story boards before the availability of previz?
In addtion to Storyboarding, which still fills an important role even on highly previsualized movies, it is possible, and still is remarkably effective although far less accurate and powerful, to do almost everything via "pantomining" with finger gestures and foam core mock ups. In the early days, when CG was still scary and new, this process was often preferred over working with computers.
Previs for 'Panic Room'.
What were some highlight projects in your career where you felt previs was making big advancements or filmmakers were taking much more advantage of its possibilities?
Fight Club is one notable example. David Fincher really embraced pre-vis fully, and we were able to prove its efficiency on a wide variety of different effects for that film, and make many good contacts. This was a really fun project, and a nice departure from slow and unglamorous motion control miniature shots. This was a real movie, a hot Director, real stars, and we were helping design real set and 'normal' movie shots. It was great fun. David is an extremely impressive guy to work for, and also really funny at times.
Previs for 'Fight Club'.
Which was probably the biggest and/or most complex job you ever worked on?
The Matrix Sequels (Reloaded and Revolutions), treated by production as one film, was by far the largest, most elaborate and in depth application of pre-vis I have witnessed. John Gaeta was the critical link here, someone with whom I had worked since the first days of pre-vis, and who wanted a truly comprehensive application of this technique. Larry and Andy Wachowski, the directors, embraced it too. The Freeway Chase from "Reloaded" is one scene I can confidently and fully state could not have been done the way it was accomplished without Pre-vis. Toy cars and pantomimed action simply cannot do what our Previs team did on that chase, and our involvement was integral at every step. The best part was the "PrePro" meeting prior to the shoot in Alameda when about 50 department heads each received a thick stack of diagrams detailing every aspect of the filmming, and to watch all of the shots backing up Larry and Andy on the big screen as they explained their goals. Kyle Robinson and Laurent Lavigne deserve high praise for pulling this all off, in addition to the rest of that strong team.
Panic Room, Fincher's next movie, was another big one. David wanted to use us as much as he could so that he would have maximal control over the shoot process. The idea was to pre-vis and diagram every shot in the film, to a NASA-like degree of accuracy, so that no unforeseen details could derail his creative process once the shoot started. He wanted to de-serialize the whole production process, so that even music and sound effects production could begin long before the shoot, and the whole movie could evolve organically at once, using the pre-vis as a base, and not one-step-at-a-time like the old days.
How does PLF prepare to work on a movie?
Well, every job is different. Generally a good first step is to meet in depth with the key people on the show, determine their priorities, and get the right kinds of relationships in place. Clarifying how pre-vis is perceived, and what it is perceived to be doing, is a critical ingredient to the success of the overall process. Then the Pre-vis Supervisor goes about assembling the asssets for the show, making important decisions about the accuracy and level of detail needed by the rest of the process. The success of the whole process is often determined by how well the Supervisor anticipates the needs of the production in this initial stage.
Do jobs get handled on site or how does the interaction work with the director or vfx supervisor?
It depends on the job. Ideally, the pre-vis is directly controlled by the Director, and managed by the VFX Supervisor to meet his or her needs, and this all happens within a domain practical to the director. Sometimes this is not possible or convenient, and the work is done remotely. But the understanding and insights that a director can get out of an "over the shoulder" session working on the raw scene directly with a pre-vis artist -- orbiting around the action, simutaneously seeing top, side, and camera views, can never be replicated. These are stressful moments, but they are the best moments in pre-vis. When filmmaking is happening via 'live' pre-vis by a director like David Fincher, it is really exciting, and super challenging to make a smooth and creatively flowing process, "live". Can be exhausting, but always rewarding.
The Matrix Reloaded.
How is previs being used directly on film sets.
The most one-to-one usage of pre-vis is in preparing Motion Control data for a shoot. The tools available inside XSI for motion editing and design are so far beyond those available through traditional MoCon tools that it really makes huge amount of sense to do as much of this design and programming in 3D, and hand off data directly.
We also have done something we call "Dur-vis" - vis that happens during the shoot. The idea is to confirm, while it is still happening, that the shoot is creating elements which will work in the context of the design for the whole sequence.This can be as simple as revising a Previs action to match how one shot was done, and then cutting the real shot into the new suquence. Pre-vis is also used by showing key crew a Virtual Stage, a fully set up version of the stage ready to show the crew, so that ideas can be tried out and tested, etc., in a "live" manner, far more quickly than having the crew move everything for real.
To what extent does PLF also handle finished work?
For years we studiously kept ourselves out of the ring of full scale VFX production, mainly strategic business decisions, which paid off, but also for a number of simple practical reasons, most of which have eventually disappeared. We now have offered full service VFX to several projects, and this is an area of the business which we are excited to keep growing. For Previs, PLF's goal is to push our design and process management abilities as far as our clients allow us, regardless of who will likely be involved in final shot production. We have taken some pre-vis to full final 2K renders, looking as close to perfect as we can achieve on tight budgets and time frames, in order to move the concept design and look development process along. Being the size we are, we are rarely the obvious choice to finish the shots we help design, but as we don't see the need for any limits on our ability to push a concept or a look further than pre-vis, if we get the change to keep pushing a design process along,we'll always try to push it one step further.
Lots of studios switched to Maya when XSI came out as it was not ready. Why did PLF stick with XSI ?
After the extreme anguish I experienced using XSI 1.0 on Panic Room, I nearly did not survive to be a user until now. Somehow we were able to make it (barely) work, and get the job done. I thought then, as now, that Softimage continues to have the best cultural perspective on the landscape of CG, and is innovating more quickly and in the direction that best serves PLF's needs. While their track record is far from perfect, I have generally felt Softimage offered me more support than I might have received from other vendors, as a long- time loyal and known customer. As Maya has increased its dominance over time, it has become more difficult to feel like this loyalty has paid off. There certainly have been moments when this loyalty was in tatters, when I was unable to really help stressed out artists get their shots to render, when the software was rightfully cursed by all at PLF, and answers from Montreal were neither helpful nor came in time.
Unlike almost any other VFX vendor, PLF is expected to share all of its data with other companies almost none of whom use XSI. So, the lengthy and error-prone process of scene conversion has not been a feather in our cap with our clients. The main reason for me is just long-term comfort and familiarity. A craftsman needs to choose his tools carefully, keep them sharp, and think of this process as an investment in the long term.
Which features of XSI do you find very useful?
Everyone has their favorite tools, so I can only speak for myself. I love the Schematic view as a way to organize my scenes. I like animating with the Mixer. I love the FX tree. I love the poly modeling tools. The animation tools are the best.
Which areas should be improved, especially also for previs?
There are lots of OGL rendering things missing, which would be fairly simple to implement, but remain missing. Such as motion blur (even of low quality), depth of field, fall-off for lights, anti alias texture maps, usable OGL toon shading. Montreal has my list (I hope!).
I think that procedural animation is the next frontier for Previs. Tools like Endorphin, which automatically animates stunts and accidents really well, need to be developed to just do normal movie action, walking and talking, etc. and even basic camera work. We need "Previs Robot".
The Matrix Reloaded.
What do you think of the current state of the film industry, if you consider it's moving more towards digital, school output tons of new digital artists and computer games are overtaking film?
I think the next step for film making is to co-opt technology from gaming into film production. Obviously, this is something everyone else has been thinking for some time, but it has not really happened yet, even at fairly basic levels. There is a large cultural and perceptual gap between people who make games and people who make films. Like the blind men and the Elephant, they understand the beast differently. I think the game industry will continue to plow ahead on the technology side, with the film industry picking their techniques and tools once they are adopted widely. Game studios will continue to use their vast resources to tap top creative talent from film and give their projects the cinematic experiential qualities which have been hard to replicate for gamers, but which are central to the language of film. Eventually some kind of hybrid game/cinema visual language will develop.
What skillset should a good previs artist have and what are the big mistakes you often see from artists trying to break into the field?
The hardest thing to find /teach is creative communication skills. Somehow, the pre-vis artist needs to communicate to a director that he or she "gets it" and can implement their vision without tedious explanation or excessive iterations. Positive and effective creative communication is the key skill a pre-vis artist needs. This cannot readily be taught. People from design school backgrounds, such as architects who are used to having their designs critiqued as part of a larger abstract design process, are good candidates. CG afficionados, who have spent their time focused on producing the final product, are less equipped to engage in this process. One thing I have found hard to teach is character animation. An artist is simply interested in animation and is therefore good at it, or they aren't and are not. This is the main area where traditional CG professionals can beat architects in doing good pre-vis.
Is there anything you would like to say to the rest of the cg community?
The main thing I would like to express to the XSI community is gratitude. Many, many people out there have contributed to PLF over our nearly 12 years, and it has been truly amazing and rewarding process. There have been so many impromptu occurrences, chance meetings and random connections that have ended up being really important to me and many other people involved in PLF. If you are one of them - Thanks!
Related Links
Pixel Liberation Front
Discuss Article at the forum
| |
|
